As of today, Cards Against Humanity is sold out. What the hell happened?
First, a little bit of recent history.
In 2012, Cards Against Humanity developed a reputation for being hard to get. We’d release a wave of games as they came in from the factory, email our list, and then sell out within days or even hours. We’d then invest almost everything back into production and go around again. This fueled the problem of resellers—a significant fraction of our games were bought by a small number of people, who would then turn around and resell them for double our price, if not higher. With each wave, resellers had more money to invest in buying a bigger and bigger fraction of our games.
This was particularly frustrating for us, since it was often hard to tell on Amazon.com that we weren’t the ones charging the ludicrous prices. We got lots of angry emails from exasperated customers demanding to know why we were price-gouging them.
Finally, this March we managed to launch with enough games to meet demand and beat the resellers, bolstered by the introduction of a US printer. We’ve been in continuous stock since then, which is great for our customers because it allows people to reliably find the game for our intended $25 price.
The holiday season is ridiculous
Staying in stock over the summer is one thing. The real challenge is to figure out how many games to make for the Christmas season. As is usually the case here at Cards Against Humanity, we had no idea what what we were doing or how to even begin to guess at that. It takes a couple of months between when we order games and when they arrive at Amazon. We had to plan well in advance. We really didn’t want to sell out before Christmas and remobilize the resellers.
We paid about $100 for an official-looking “toy and games industry report” to find out how sales generally looked in in November and December compared to the summer months. Compared to July sales, the report said that November and December sales were about two and three times higher, respectively. This seemed like a really simplistic approach. It didn’t account for growth as more people heard about and played the game. It didn’t account for the planned launch of a new expansion. And it didn’t account for the effects of any of our efforts to surprise and delight our fans with risky stunts like our 12 Days of Holiday Bullshit.
More importantly, those numbers didn’t say anything about how uncertain they were. Even if we believed that the report had the right answer on average, we wanted a less than 50% chance of selling out before Christmas. How many games should we make to achieve that?
The wisdom and uncertainty of the crowd
Perhaps the simplest way to figure out the range of likely holiday sales figures is to poll the group. There’s eight of us who make the game so that means we’ll get eight different answers. In theory, the range of our answers should give us an estimate of how uncertain we are about them. This approach would utilize the wisdom of the crowd, but it would be vulnerable to our own biases. Guessing total holiday sales is really hard and we might all over- or underestimate them because of the similarities of our experiences.
Instead of trying to guess the result of a large number of uncertain factors, we tried to break the problem into simpler components that we could more easily guess. We constructed a mathematical model for our sales with a number of inputs like “how fast is Cards Against Humanity growing,” “how much better would December be than July if we weren’t growing,” and “how much will launching the fourth expansion help the sales of our other stuff.” Given a value for each of these dozen or so inputs (and about a dozen more related to expansions and the Bigger, Blacker Box), the model predicted how many games we’d sell each month.
What we really wanted was an estimate of how uncertain we were about our future sales, not just the prediction of a complicated model. So we did a little experiment. We assumed that every input to our model was a random number drawn from a Bell curve or “normal distribution.” The average value and the spread in values of each input number reflected the average and spread of the guesses of the CAH creators. We picked a set of random inputs in this way, computed the expected sales each month using our mathematical model, and then it again and again. 100,000 times, in fact. In the end, we got a good sense of the range of predicted values and how likely each was based on our guesses and the spread in our guesses (see above).
The method has a fancy name, a “Monte Carlo.” It isn’t so different from what Nate Silver did for his forecast of the 2012 election, though not nearly as sophisticated. The basic idea is that we translated our uncertainty about a large number of relatively concrete inputs into one abstract and hard-to-guess output: our total sales during the holiday season.
Why we were wrong
What we didn’t expect was how truly insane the holiday season is in this country. As a group, we guessed that our December sales would be about four times higher than our July sales (remember, the industry report we bought said holiday sales were three times higher at most). While we slightly overestimated sales for September, October, and November, we REALLY underpredicted December sales, which wound up being seven times higher than July.
Why did that happen? Well, the truth is, we have no idea, but if we had to guess, it’s because some of the crazy stunts we did around the holidays like our Black Friday “everything costs $5 more” promotion or our $100,000 Donor’s Choose Shopping Spree seem to have been successful. We also got some unexpected press, like being added to some nice end-of-year gift guides. Cards Against Humanity also sells well when lots of people are playing the game and introducing it to friends, which happens most often when people are home avoiding their families.
The bad news here is that we’ve sold out for the holidays, and we’re going to be back to watching resellers jack up the price for the next few weeks, until we get new games in stock.
The good news is that we were wrong in an interesting way, learned from our mistakes, and get to share that lesson with you.